J Mol Med DOI 10.1007/s00109-012-0936-6

ORIGINAL ARTICLE

LGALS3BP, lectin galactoside-binding soluble 3 binding protein, induces vascular endothelial growth factor in human breast cancer cells and promotes angiogenesis

Enza Piccolo & Nicola Tinari & Daniela Semeraro & Sara Traini & Imma Fichera & Albana Cumashi & Rossana La Sorda & Francesca Spinella & Anna Bagnato & Rossano Lattanzio & Maurizia D’Egidio & Annalisa Di Risio & Pavlos Stampolidis & Mauro Piantelli & Clara Natoli & Axel Ullrich & Stefano Iacobelli

Received: 3 May 2012 /Revised: 6 July 2012 /Accepted: 17 July 2012 # Springer-Verlag 2012

Abstract Elevated serum or tissue levels of lectin galactoside- migration and, more importantly, a reduced expression of binding soluble 3 binding protein (LGALS3BP) have been vascular endothelial growth factor (VEGF). Production of associated with short survival and development of metastasis VEGF, that was restored by exposure of silenced cells to in a variety of human cancers. However, the role of recombinant LGALS3BP, required an intact PI3k/Akt signal- LGALS3BP, particularly in the context of tumor–host relation- ing. Furthermore, we show that LGALS3BP was able to ships, is still missing. Here, we show that LGALS3BP knock- directly stimulate HUVEC tubulogenesis in a VEGF- down in MDA-MB-231 human breast cancer cells leads to a independent, galectin-3-dependent manner. Immunohisto- decreased adhesion to fibronectin, a reduced transendothelial chemical analysis of human breast cancer tissues revealed a correlation among LGALS3BP expression, VEGF expression, and blood vessel density. We propose that in addition to its prometastatic role, LGALS3BP secreted by breast cancer cells Electronic supplementary material The online version of this article (doi:10.1007/s00109-012-0936-6) contains supplementary material, functions critically as a pro-angiogenic factor through a dual which is available to authorized users. mechanism, i.e by induction of tumor VEGF and stimulation of endothelial cell tubulogenesis. E. Piccolo : N. Tinari : S. Traini : R. La Sorda : M. D’Egidio : A. Di Risio : M. Piantelli : C. Natoli : S. Iacobelli MediaPharma s.r.l., Keywords Angiogenesis . VEGF . LGALS3BP . Via Colle dell’Ara, Extracellular matrix . Galectin-3 Chieti, Italy

E. Piccolo (*) : N. Tinari : D. Semeraro : S. Traini : I. Fichera : : : : ’ : A. Cumashi: R. La Sorda: R. Lattanzio: M. D Egidio A. Di Risio M. Piantelli C. Natoli S. Iacobelli Introduction Department of Biomedical Sciences, “G. d’Annunzio” University and Foundation, Via dei Vestini, 66100 Chieti, Italy The development of human cancer is a multistep pro- e-mail: [email protected] cess characterized by the accumulation of progressive : genetic alterations that confer to tumor cells the ability F. Spinella A. Bagnato to survive, proliferate, and metastasize. Once they have Laboratory of Molecular Pathology, Regina Elena National Cancer Institute, expanded to a critical level, tumor cells find a way to Rome, Italy promote new vasculature development, a process known : as tumor angiogenesis in order to progress and metas- P. Stampolidis A. Ullrich tasize [1]. This "angiogenic switch" can occur at differ- Department of Molecular Biology, Max Planck Institute of Biochemistry, ent stages of tumor progression and is dictated by an Martinsried/Munich, Germany imbalance between factors that either induce or oppose J Mol Med angiogenesis in favor of the former [2]. Vascular endo- heat-inactivated FBS supplemented with EGM-2 Single- thelial growth factor (VEGF) represents the prototypical Quots (Lonza, Basel, Switzerland). pro-angiogenic factor [3]. Although hypoxia appears to For transfection of HUVEC, siRNA duplexes directed be the main driver of tumor angiogenesis, growth fac- against human galectin-3 or nontargeting siRNA duplex or tors and inflammation are able to boost VEGF produc- siRNA duplexes against human VEGF (Thermo Scientific, tion through paracrine or autocrine mechanisms [4, 5]. Waltham, MA, USA) were used according to the manufac- However, much more remains to be understood about turer’s instructions. SiRNA duplexes against LGALS3BP the molecular nature of unidentified factors that are also were obtained from Qiagen (Qiagen, Hilden, Germany). believed to have the ability to promote angiogenesis in For collection of the conditioned media (CM), cells were the development of a variety of human cancers. grown until 70 % confluent. The next day, the medium was LGALS3BP (also known as 90 K or Mac-2 BP) is a large replaced with serum-free medium after washing with oligomeric glycoprotein composed of 90-kDa subunits that phosphate-buffered saline (PBS). The medium was collected was originally identified as a tumor-secreted antigen [6]andas 24–48 h later and centrifuged at 1,200 rpm for 15 min to a ligand of the lactose-specific S-type lectin, galectin-3 (for- remove cells and debris. merly Mac-2) [7]. Over the past decade, considerable attention has been focused on a potential role of LGALS3BP in the LGALS3BP gene knockdown development of human cancer. Immunohistochemical and gene expression analysis showed significantly higher levels A 21-nucleotide sequence corresponding to nucleotide of LGALS3BP in different types of human malignancies [8]. 2216-2236 of human LGALS3BP mRNA (NCBI Acces- Furthermore, clinical studies have revealed that elevated se- sion NM-005567.3) or a 21-nucleotide sequence with no rum or tumor tissue levels of LGALS3BP are associated with significant homology to any mammalian gene sequence a shorter survival in patients with breast carcinoma [9, 10], serving as a non-silencing control (OligoEngine, Hercules, lymphoma [11], pleural mesothelioma [12], and non-small CA, USA) was inserted into the pSUPER.retro.puro (Oli- cell lung carcinoma [13]. goEngine, Hercules, CA, USA). After transformation into Taken together, these data indicate that expression levels DH5α competent cells (Invitrogen, Carlsbad, CA, USA), of LGALS3BP may serve as a prognostic indicator. None- the recombinant plasmids were confirmed by PCR ampli- theless, little is known regarding LGALS3BP activity in fication, restriction enzymes digestion, and DNA sequenc- human cancers. ing. The generation of knockdown cells was performed as In the current study, we tested the hypothesis that previously described [14]. LGALS3BP acts as a pro-angiogenic factor. We find and report for the first time that LGALS3BP induces VEGF ex- pression in human breast cancer cells by activation of the Production of recombinant LGALS3BP PI3k/Akt pathway. Furthermore, we show that LGALS3BP directly acts on endothelial cells, promoting angiogenesis in Recombinant LGALS3BP was immunoaffinity-purified from vitro and in vivo in a VEGF-independent, galectin-3- free supernatant of human embryonic kidney EBNA-293 cells dependent manner. (Invitrogen) [15] transfected with LGALS3BP cDNA [16].

Quantitative real-time PCR Materials and methods Total RNA was extracted using the RNeasy mini kit Cell lines and culture (Qiagen), quantified by optical density, and checked for quality by gel electrophoresis. RNA was reverse- The estrogen receptor (ER)-negative human breast cancer transcribed with cDNA High Capacity kit (Applied Bio- cell lines MDA-MB 231 and SK-BR-3 and the ER-positive system, Foster City, CA, USA) according to the manu- MCF-7 and T47D were purchased from American Type facturer's instructions. Quantitative real-time PCR was Culture Collection (Rockville, MD, USA). Cells were main- performed using 96-well optic plates on an ABI Prism tained in Dulbecco’s modified Eagle’s medium (Invitrogen, 7500 Sequence Detection System (Applied Biosystems). Carlsbad, CA, USA) with 10 % heat-inactivated fetal bovine ThepredevelopedTaqManassayreagentsofhuman serum (FBS; Invitrogen, Carlsbad, CA, USA), L-glutamine, VEGF (ID: Hs00173626_ m1) and human LGALS3BP and antibiotics (Sigma Aldrich Corporation, St. Louis, MO, (ID: Hs00174774_ m1) were used according to the man- USA). The human umbilical vein endothelial cells ufacturer's instructions. Relative expression values were (HUVEC) were purchased from Lonza (Lonza, Basel, Swit- obtained using the Q-Gene software and normalized to zerland) and maintained in EBM-2 medium containing 10 % the expression of GAPDH (ID: Hs0017266705g1). J Mol Med

Confocal microscopy and spreading assay microscope and counted from ten random fields of ×200 magnification. MDA-MB-231 cells and HUVEC grown on coverslips were treated as indicated, fixed in 4 % paraformaldehyde for Enzyme-linked immunosorbent assay 15 min at room temperature, permeabilized with 0.25 % Triton X-100 for 5 min, and blocked with 0.1 % bovine Sandwich enzyme-linked immunosorbent assay (ELISA) was serum albumin for 1 h at room temperature. Coverslips were performed using kits specific for VEGF-A (Invitrogen) and then incubated for 2 h at room temperature with the indicat- LGALS3BP (Diesse, Siena, Italy) in accordance with the ed primary antibodies, followed by Alexa Fluor specific manufacturers’ instructions. secondary antibodies (Molecular Probes, Life Technologies, Paisley, UK). TRITC-labeled phalloidin was used to visual- Tubulogenesis assay ize actin cytoskeleton; DRAQ5 (Vinci-Biochem, Firenze, Italy) was used to visualize nuclei. Images were acquired Capillary tube formation (tubulogenesis) assay was per- with the Zeiss LSM 510 meta-confocal microscope (Zeiss, formed as described [18] with a few modifications. HUVEC Oberkochen, Germany) using 488-, 543-, and 633-nm were resuspended in a serum-free EBM-2 or the indicated lasers. CMs and seeded in Matrigel-coated chamber slides at a 4 density of 5×10 cells/well. Human recombinant VEGF165 Fluorescent labeling of MDA-MB-231 Cells (R&D Systems, Minneapolis, MN, USA) was used as pos- itive control. After incubation at 37 °C for 4 h, cultures were Serum-starved MDA-MB-231 cells were dispersed in photographed. For each individual well, three photographs 0.25 % trypsin, resuspended in serum-free medium, and were taken at different locations and analyzed by the Image- labeled by incubation with 5 μM calcein AM (Molecular Pro Plus software (Media Cybernetics, Silver Spring, MD, Probes) for 30 min. Cells were washed twice with PBS and USA). resuspended in serum-free medium at a concentration of 5×106 cells/mL. Calcein-labeled cells were then used for In vivo Matrigel plug assay adhesion and transmigration assays. The Matrigel plug assay was performed as described previ- ously [19] with some modifications. Female athymic (nu+/ Adhesion assay nu+) mice (Charles River Laboratories, Milan, Italy) were subcutaneously injected with 0.5 mL Cultrex (basic mem- Human MDA-MB-231 cells were serum starved for 18 h. brane extract, BME, Trevigen, Gaithersburg, MD, USA) Cells were washed with PBS (containing 0.1 % BSA), containing 5×106 shRNA control or shRNA LGALS3BP gently detached with 0.25 % trypsin, and plated onto 96- cells. Human recombinant VEGF (R&D Systems) and μ 165 well culture plate precoated with 10 g/mL of BSA, PBS were used as positive and negative control, respective- collagen IV (Millipore, Billerica, MA, USA), collagen ly. Six animals per group were used. Seven days after V (Sigma Aldrich, St. Louis, MO, USA), fibronectin injection, mice were sacrificed, and plugs were removed (Fn) (Millipore, Billerica, MA, USA), and laminin (Becton and photographed. Hemoglobin content of the plug was Dickinson, Franklin Lakes, NJ, USA). After incubation and assayed using the Drabkin method [20] (Drabkin reagent washing, the adherent cells were dyed with 0.1 % crystal kit 525; Sigma Aldrich, St. Louis, MO, USA). Hemoglobin violet for 30 min at room temperature and solubilized with content was normalized to the plug weight and the values μ 150 L/well of 2 % SDS. Absorbance at 550 nm was expressed as fold increase compared to the control. The measured with a microplate reader (SpectraMax, Molecular procedure involving animals and their care was conducted Devices, Sunnyvale, CA, USA). according to the institutional guidelines in compliance with Adhesion assay of MDA-MB-231 cells to HUVEC national and international laws and policies. monolayer was performed as described [17]. Attached tu- mor cells were observed under a fluorescent microscope and Western blotting counted from ten random fields of ×200 magnification. Cell lysates were subjected to 10 % SDS–PAGE and West- Transendothelial penetration of MDA-MB-231 cells ern blotting using the following antibodies: rabbit anti-Ser (473)-Akt, rabbit anti-phospo-p42/44 Mapk (ERKs), rabbit Transendothelial penetration (TEP) of MDA-MB-231 cells Akt, rabbit anti p42/44 Mapk (all diluted 1:1,000; all from on HUVEC cells was performed as described [17]. Trans- Cell Signaling Technology, Danvers, MA, USA), mouse migrated tumor cells were observed under a fluorescent anti-HIF-1 α (1:1,000; Becton Dickinson), and mouse J Mol Med anti-actin (1:3,000; Sigma Aldrich). Specific signals were stable short hairpin RNA (shRNA). LGALS3BP-specific detected using an ECL kit. shRNA reduced the expression of LGALS3BP mRNA and protein by 80 to 90 % (data not shown). Similar results were Immunohistochemistry obtained by silencing LGALS3BP in three other human breast cancer cell lines, SK-BR-3, MCF-7, and T47D (data not The study of breast cancer samples was approved by “G. shown). To test the effect of LGALS3BP knockdown on cell D’Annunzio” University and Foundation Local Ethics Com- adhesion, cells were seeded on plates coated with collagen IV, mittee. Archival, paraffin-embedded breast cancer tissues collagen V,fibronectin, and laminin and the extent of adhesion were collected from 137 female patients (118 ductal and quantified by staining cells with crystal violet. As compared to 19 lobular carcinoma) over the period from October 2007 to shRNA control cells, adhesion on fibronectin was significant- April 2010. Immunostaining was performed with the fol- ly reduced in shRNA LGALS3BP cells (Fig. 1a). Image lowing primary antibodies: mouse monoclonal antibody acquisition with laser scanning microscopy revealed that 1A4. 22 [21] for LGALS3BP; clone JH121 (NeoMarkers, LGALS3BP knockdown cells appeared more round and less Fremont, CA, USA) for VEGF, and clone 1A10 (Ylem, spread when plated on fibronectin, with few adhesion contacts Rome, Italy) for CD31/PECAM. Sections for LGALS3BP as revealed by staining for focal adhesion kinase (FAK; and VEGF immunolabeling were subjected to antigen re- Fig. 1b), suggesting an involvement of LGALS3BP in cyto- trieval by treating them with microwave (10 min) in citrate skeleton rearrangement after adhesion to fibronectin. On buffer, pH 6.0 and EDTA buffer, pH 8.5. For CD31/PEAM, Western blot, shRNA LGALS3BP cells plated on fibronectin antigen retrieval was carried out by immersing sections in showed a reduced activation of FAK and Akt as compared to boiling citrate buffer (pH 6.0) for 1 min in a domestic shRNA control cells, whereas ERK activation was unchanged pressure cooker before cooling in water. Positive controls (Fig. 1c). These data indicate that LGALS3BP knockdown were stained in parallel. Omission of the primary antibody impairs adhesion and spreading on fibronectin of MDA-MB- was used as a negative control. Staining of LGALS3BP and 231 cells, possibly through a reduced activation of FAK and VEGF was expressed as the percentage of positive staining Akt. cells. The microvessel density was determined by micro- To assess the involvement of LGALS3BP in cell motility, scopic evaluation of three different fields under ×200 mag- we performed a transwell assay. A similar fraction of cells nification. The stained vessels from the three sites were then passed through the filter, independently of whether they counted and averaged. A single pathologist (M.P.) selected were knocked down for LGALS3BP or not, exposed to the areas of tumor with the highest number of vessels. Areas CM of silenced or control cells, or exposed to recombinant of sclerosis or dense inflammation were avoided. LGALS3BP (data not shown).

Statistical methods Knockdown of LGALS3BP impairs adhesion to endothelial cells and transendothelial penetration To test differences in means of cell numbers and biological activities, a two-sided t test was used. The correlation To study the involvement of LGALS3BP in the adhesion of among expression of LGALS3BP, VEGF, and CD31/ tumor cells to endothelium and their intravasation, calcein- PECAM was evaluated by the Pearson Chi-square. High labeled cells were added to HUVEC monolayers and the and low categories were obtained using the following me- number of adherent cells assessed under a fluorescent mi- dian cutoff values: for LGALS3BP, 40 %; for VEGF, 61 %; croscope. In serum-free conditions, no significant difference for CD31/PECAM, 10 vessels. The SPSS version 15.0 (SPSS, was observed between shRNA control and shRNA Chicago, IL, USA) was used throughout, and P<0.05 was LGALS3BP cells (Fig. 2a). On the other hand, when the considered statistically significant. assay was performed in the presence of CM from shRNA control cells, adhesion increased if compared to that ob- served in serum-free conditions, whereas it returned to the Results basal level in the presence of CM from shRNA LGALS3BP cells. Adhesion was not modified by the addition of recom- Knockdown of LGALS3BP impairs cell adhesion binant LGALS3BP (10 μg/mL; Fig. 2a). and spreading TEP of tumor cells was next explored. To this end, calcein- labeled MDA-MB-231 cells were added to HUVEC mono- To gain further insight into the mechanisms of LGALS3BP- layers in Transwell apical chambers and the number of pene- dependent cell adhesion, expression of endogenous trating cells assessed under a fluorescent microscope. In the LGALS3BP was reduced in the human breast cancer cell line presence of serum, no difference in TEP was seen between MDA-MB-231 by pSuper retro-based vectors to express shRNA control and shRNA LGALS3BP cells (Fig. 2b). The J Mol Med

Fig. 1 shRNA-mediated inhibition of LGALS3BP expression impairs plated on PLL or Fn-coated well for 20 min, lysed, and analyzed for cell adhesion and spreading. a Adhesion assay of shRNA control and the indicated proteins. Cells maintained in suspension (Susp) were used LGALS3BP MDA-MB-231 cells on ECM proteins. *P<0.05 as com- as control. Fold increase was calculated by densitometric scanning of pared to shRNA control cells. b Confocal microscopy images showing the bands using ImageJ. Ratio between the level of the phosphorylated shRNA control and shRNA LGALS3BP MDA-MB-231 cells seeded protein and the corresponding total protein level was indicated. The on poly-L-lysine (PLL) or fibronectin (Fn) for 20 min. Pictures show ratio of shRNA control cells was arbitrarily set to 1. Data are repre- FAK (green), actin (red), and nuclei (blue). c Western blotting of sented as mean±SEM of at least three independent experiments shRNA control and LGALS3BP MDA-MB-231 cells. Cells were presence of CM from shRNA control cells led to an increased Exposure of shRNA LGALS3BP cells to the indicated con- TEP of tumor cells that reached the level of that observed in centration of LGALS3BP for 6 h significantly increased thepresenceofserum(Fig.2b); CM from shRNA mRNA expression of VEGF (Fig. 3a). The induced transcript LGALS3BP cells did not affect transmigration. As expected, was functional, as demonstrated by an increased secretion of VEGF, used as a positive control, led to an increased TEP, VEGF protein into the culture medium (Fig. 3b). The hypoxia whereas recombinant LGALS3BP failed to increase TEP. inducer, cobalt chloride (CoCl2), was used as a positive control Expression of VEGF in MDA-MB-231 shRNA LGALS3BP for VEGF stimulation. Similar results were obtained in SK- cells was reduced by approximately 50 % both at mRNA and BR-3,MCF-7,andT47Dcells(Fig.S2). protein levels (Fig. 2c, d). Similar results were obtained in SK- To explore the possible pathway through which BR-3, MCF-7 and T47D cells (Fig. S1). LGALS3BP induces VEGF, we first focused on HIF-1α,a major regulator of VEGF expression. To this end, shRNA LGALS3BP stimulates VEGF production via PI3k/Akt control and shRNA LGALS3BP cells were exposed to pathway in human breast cancer cells LGALS3BP (10 μg/mL) for different times and HIF-1α levels analyzed by Western blotting. As shown in Fig. 3c,HIF-1α Given the observed downregulation of VEGF in shRNA expression was not influenced by LGALS3BP treatment, LGALS3BP cells, we wanted to determine whether whereas it was clearly induced after treatment of cells with

LGALS3BP could directly stimulate VEGF production. CoCl2 used as a positive control. J Mol Med

Fig. 2 shRNA-mediated inhibition of LGALS3BP expression impairs cells, VEGF (100 ng/mL), or human recombinant LGALS3BP (10 μg/ adhesion and penetration of MDA-MB-231 cells across HUVEC. a mL) were added to the basolateral chambers. After 6 h of incubation, Adhesion assay of MDA-MB-231 cell clones on HUVEC monolayers. the apical chamber was washed extensively with PBS. The number of Calcein-labeled cells were added to each well in serum-free medium, migrating tumor cells was evaluated by a fluorescent microscope in ten CM medium from shRNA control and shRNA LGALS3BP cells, randomly selected fields at ×200 magnification. The results are pre- VEGF (100 ng/mL), or recombinant LGALS3BP (10 μg/mL) and sented as the mean±SD of duplicate samples and are representative of incubated for 2 h. The number of attached tumor cells was evaluated three individual experiments. *P<0.005 between serum free and the in ten randomly selected fields at ×200 magnification. The results are CM of shRNA control cells. **P<0.001 between CM of shRNA presented as the mean±SD of triplicate samples. *P<0.005 between control and shRNA LGALS3BP cells. c Levels of VEGF mRNA and serum free and the CM of shRNA control cells. **P<0.001 between d VEGF protein in shRNA control and shRNA LGALS3BP MDA- CM of shRNA control and shRNA LGALS3BP cells. b Trasmigration MB-231 cells were measured by quantitative real-time PCR and of calcein-labeled MDA-MB-231 clones through HUVEC monolayers ELISA, respectively. *P<0.05 compared to shRNA control cells. Data grown on fibronectin-coated 24-well transculture inserts. Serum-free are represented as mean±SEM of at least three independent medium, CM medium from shRNA control and shRNA LGALS3BP experiments

We then analyzed other pathways involved in VEGF pro- LGALS3BP silencing inhibits in vitro and in vivo duction. As shown in Fig. 3c, exposure of cells to LGALS3BP angiogenesis (10 ug/mL) resulted in a time-dependent increase of Akt phosphorylation at Ser473 residue, with a maximal response To determine whether the reduced VEGF expression between 2 and 10 min. No effect of LGALS3BP on ERKs was induced by silencing LGALS3BP gene is biologically observed. Interestingly, in cells incubated with LGALS3BP, relevant, CMs from shRNA control and shRNA VEGF secretion was enhanced up to 1.5-fold as compared to LGALS3BP cells were tested for their ability to affect unstimulated cells (1,990±11 vs 1,343±37 pg/mL in ctrl); co- in vitro HUVEC tubulogenesis. Consistent with the re- incubation with the PI3k/Akt inhibitor, wortmannin, com- duced content of VEGF, HUVEC exposed to CM from pletely blocked LGALS3BP-induced increase in VEGF se- shRNA LGALS3BP cells produced less tube-like struc- cretion (1,990±11 vs 590±29 pg/mL; Fig. 3d). Taken tures as compared to those exposed to CM from shRNA together, these results strongly suggest that increased secretion control cells (Fig. 4a, upper panel). These changes were of VEGF by MDA-MB-231 cells in response to LGALS3BP quantified and found to be significant (Fig. 4a,lower is mediated by the PI3k/Akt pathway. panel). J Mol Med

Fig. 3 LGALS3BP stimulates VEGF production by MDA-MB-231 After treatment, cell lysates were harvested, and samples were sub- cells via PI3k/Akt pathway. shRNA LGALS3BP cells were exposed jected to Western blotting to detect the indicated protein. Fold increase for 24 h with 5 and 10 μg/mL of LGALS3BP or 200 μg/mL CoCl2 as a was calculated by densitometric scanning of the bands using ImageJ. positive control. a Total RNAs were prepared, and VEGF mRNA Ratio between the level of the phosphorylated protein and the levels were determined by real-time PCR using VEGF-A-specific corresponding total protein level is shown. The ratio of cells not primers. The VEGF mRNA arbitrary unit for shRNA control cells exposed to LGALS3BP was arbitrarily set to 1. d The amount of was 1.6. *P<0.05 compared to cells not exposed to LGALS3BP. b VEGF secreted in CM of cells incubated for 24 h with LGALS3BP The amounts of VEGF in CM of shRNA LGALS3BP cells were at 5 and 10 μg/mL with or without a co-incubation with the PI3k/AKT quantified with VEGF Human ELISA Kit. The amounts were normal- inhibitor, wortmannin (100 nM), was quantified by the VEGF Human ized for the number of cells at the time of media collection. Data are ELISA Kit and normalized for the number of cells at the time of media represented as mean±SEM of at least three independent experiments. collection. *P<0.05 compared to cells not exposed to LGALS3BP. *P<0.05 compared to cells not exposed to LGALS3BP. c Western **P<0.001 compared to LGALS3BP stimulated cells. Data are repre- blotting of shRNA LGALS3BP MDA-MB-231 cells treated with sented as mean±SEM of at least three independent experiments LGALS3BP (10 μg/mL) or CoCl2 (200 μg/mL) for the indicated times.

Next, we questioned whether LGALS3BP gene knock- plugs containing shRNA control cells. Quantification of the down could influence tumor angiogenesis in vivo. To this angiogenic response by determination of the hemoglobin con- end, nude mice were injected subcutaneously with shRNA tent showed that plugs containing shRNA control cells had a control or shRNA LGALS3BP MDA-MB-231 cells sus- significantly higher (3.4-fold) level of vascularization than pended in growth factor-reduced Matrigel. Injections of Matri- those containing shRNA LGALS3BP MDA-MB-231 cells. gel containing PBS or VEGF were used as negative and (Fig. 4b, lower panel). These data demonstrated that positive control, respectively. As shown in Fig. 4b, upper LGALS3BP has the capacity to induce the formation of new panel, plugs containing shRNA LGALS3BP cells exhibited blood vessel in vivo, and support the hypothesis that a reduced angiogenic response compared to that observed in LGALS3BP may function as a pro-angiogenic factor. J Mol Med

Fig. 4 LGALS3BP silencing inhibits in vitro and in vivo angiogene- CM of shRNA control cells. Scale bars, 100 μm. b Matrigel plug assay sis. a HUVEC tubulogenesis in vitro. Upper panel, representative of shRNA control or shRNA LGALS3BP MDA MB-231 cells. Upper phase-contrast photographs of capillary-like tube formation of HUVEC panel, representative Matrigel plugs containing PBS, shRNA control on Cultrex incubated with EBM-2 alone, CMs of shRNA control, or cells, shRNA LGALS3BP cells, or VEGF. Lower panel, quantification shRNA LGALS3BP MDA MB-231 cells. VEGF (50 ng/mL) was used of novel blood vessel formation through measurement of the hemoglo- as a positive control. Lower panel, quantification of tube formation was bin content of the plugs. Six mice per group of treatment were used. *P performed by counting closed areas (tubes) in four different fields. <0.004 compared to PBS; **P<0.05 compared to shRNA control Data are represented as mean±SEM from at least three independent cells. Data are represented as mean±SEM of at least three independent experiments. *P<0.01 compared to EBM-2; **P<0.05 compared to experiments

Recombinant LGALS3BP stimulates endothelial cell CMs of the different silenced clones added to HUVEC to angiogenesis in a VEGF-independent, galectin-3-dependent perform tube formation. As expected, in the presence of CM manner from siRNA VEGF or siRNA LGALS3BP cells, tube for- mation was reduced by about 30 % as compared to that seen Above, we have shown that HUVEC exposed to CM from with CM from siRNA control cells (Fig. 5a). The use of CM shRNA LGALS3BP cells form less tube-like structures, from LGALS3BP/VEGF double knocked down cells probably as a result of a reduced VEGF content. However, resulted in a greater reduction of tube formation. Moreover, a direct involvement of LGALS3BP needs to be ruled out. addition of recombinant LGALS3BP to CM of double To this end, MDA-MB-231 cells were knocked down for knocked down cells completely rescued tube formation LGALS3BP and/or VEGF by using specific siRNA and the (Fig. 5a), indicating a VEGF-independent effect of J Mol Med

Fig. 5 Recombinant LGALS3BP stimulates endothelial cell angiogen- compared to CM of siRNA LGALS3BP; ##P<0.0001 compared to esis in a VEGF-independent manner. a HUVEC tubulogenesis in vitro. CM of siRNA LGALS3BP/VEGF). b HUVEC tubulogenesis in vitro. Upper panel, representative phase-contrast photographs of capillary- Upper panel, representative phase-contrast photographs of capillary-like like tube formation by HUVEC on Cultrex incubated with serum-free tube formation of HUVEC on Cultrex, incubated with serum-free EBM-2 EBM-2, CMs from siRNA control, siRNAVEGF, siRNA LGALS3BP, in the presence of LGALS3BP (10 μg/mL) or VEGF (50 ng/mL). To test or siRNA VEGF/LGALS3BP MDA-MB-231 cells. VEGF (50 ng/mL) the effect of a LGALS3BP blocking antibody, LGALS3BP was incubated was used as a positive control. LGALS3BP (10 μg/mL) was added to with the indicated concentration of SP-2 prior to the assay. Scale bars, the CM of siRNAVEGF/LGALS3BP cells. Scale bars, 100 μm. Lower 100 μm. Lower panel, quantification of tube formation was performed by panel, quantification of tube formation was performed by counting counting the number of closed areas (tubes) in four different fields. Data closed areas (tubes) in four different fields. Data are collected from at are collected from at least three independent experiments. (*P<0.001 least three independent experiments. (*P<0.001 compared to EBM-2; compared to cells maintained in serum-free EBM-2; **P<0.005 com- **P<0.008 compared to CM of siRNA control cells; #P<0.005 pared to LGALS3BP alone)

LGALS3BP. To confirm that the effect was specific for whether the lectin is a necessary ligand for LGALS3BP- LGALS3BP, we used an anti-LGALS3BP blocking anti- induced tubulogenesis. To this end, tube formation assay body, SP-2, which was shown to inhibit LGALS3BP func- was performed using HUVEC knocked down for galectin-3. tion (30). SP-2 dose dependently inhibited tube formation As expected, galectin-3 silenced cells produced a less tube- (Fig. 5b, upper panel); the effect was quantified and found to like structure as compared to control cells (Fig. S3A). Most be significant (Fig. 5b, lower panel), demonstrating a importantly, stimulation of tubulogenesis by LGALS3BP LGALS3BP specific role in tube formation. was lost in galectin-3 knocked down cells (Fig. S3A). Fur- Recently, data have been presented to show that galectin- thermore, the use of a galectin-3 blocking antibody, M3-38, 3 is required for VEGF and bFGF-mediated angiogenic inhibited LGALS3BP-induced tube formation. Finally, response in endothelial cells [22]. Since galectin-3 is a LGALS3BP was able to induce FAK phosphorylation at well-known ligand of LGALS3BP, we wanted to verify tyrosine 125 (Fig S3B) and its translocation to the plasma J Mol Med membrane to distinct areas of the plasma membrane, to as an effective therapeutic modality still awaits the complete constitute the so-called focal adhesion (Figure S3C) [22]. understanding of the many biochemical and molecular signals Altogether, the results indicate that galectin-3 is the ligand that regulate the process on one hand and the identification of responsible for LGALS3BP-stimulated tubulogenesis of en- inducers and inhibitors of angiogenesis on the other. dothelial cells. LGALS3BP is overexpressed in many cancers, and high levels are associated with distant metastasis and poor survival LGALS3BP is associated with VEGF expression and blood in patients with a variety of cancers. In the present paper, we vessel density in breast cancer show for the first time that LGALS3BP induces VEGF in breast cancer cells and promotes endothelial angiogenesis To further substantiate a link between LGALS3BP and both in vitro and in vivo (Figs. 3 and 4). The pro-angiogenic angiogenesis, we performed immunohistochemistry to stain effect of LGALS3BP appears particularly interesting in light LGALS3BP, CD31/PECAM, a marker of human vascular of the observed positive correlation between the expression endothelial cells, and VEGF in 137 cases of human breast levels of LGALS3BP, VEGF, and blood vessel density in carcinoma. We found that cancers displaying a high percent- human breast cancer (Fig. 6). Recently, VEGF and its recep- age of LGALS3BP-positive cells (Fig. 6a) exhibited a high tors were also involved in promoting tumor metastasis [25]. degree of VEGF expression (Fig. 6c) and vascularization These findings could explain the phenomena reported previ- (Fig. 6e). On the contrary, cancers with low LGALS3BP ously that breast cancer patients with elevated serum or tumor (Fig. 6b) showed low VEGF expression (Fig. 6d) and CD- levels of LGALS3BP have a significantly higher propensity to 31/PECAM density (Fig. 6f). Pearson Chi-square analysis develop metastases and a shorter survival [9, 10]. showed that LGALS3BP expression directly correlated with The mechanism(s) through which LGALS3BP induce(s) VEGF expression (P<0.009) and blood vessel density (P< VEGF in breast cancer cells is (are) presently unknown. 0.0001; not shown). This induction is dependent on an intact PI3k/Akt signaling pathway as it was completely blocked by a PI3k/Akt inhib- itor (Fig. 3d). Several reports have documented that PI3k/ Discussion Akt signaling is required for VEGF expression through HIF- 1α in response to growth factor stimulation and oncogene Tumor-associated angiogenesis is a hallmark of a growing activation [26, 27]; however, the PI3k/Akt activation was tumor and thus serves as a rational target for cancer eradica- also associated to VEGF increase with a HIF-1α- tion [23, 24]. However, blocking the process of angiogenesis independent mechanism [28]. The PI3k/Akt network plays

Fig. 6 Expression levels of LGALS3BP correlate with VEGF PECAM by immunohistochemistry. Representative photographs of expression and blood vessel density in human breast cancer. high (a) and low/absent (b) immunostaining for LGALS3BP, One hundred and thirty-seven infiltrating breast carcinomas were VEGF (c, d), and CD-31 positive vessels (e, f). Original magni- analyzed for the expression of LGALS3BP, VEGF, and CD-31/ fication ×40 (scale bar,20μm) J Mol Med a key regulatory function in cell survival, proliferation, and References angiogenesis; genetic aberrations found at different levels, either with activation of oncogenes or inactivation of tumor 1. Folkman J (1971) Tumor angiogenesis: therapeutic implications. N suppressors, make this pathway one of the most commonly Engl J Med 285:1182–1186 disrupted in human breast cancer. 2. Baeriswyl V, Christofori G (2009) The angiogenic switch in car- We add new insight in the poorly understood mechanism cinogenesis. Semin Cancer Biol 19:329–337 underlying the promoting effect of LGALS3BP on HUVEC 3. Ferrara N, Henzel WJ (1989) Pituitary follicular cells secrete a novel heparin-binding growth factor specific for vascular endothe- tubulogenesis providing that this effect appears to be VEGF- lial cells. Biochem Biophys Res Commun 161:851–858 independent, as it was seen after exposure to CM from 4. Cohen T, Nahari D, Cerem LW, Neufeld G, Levi BZ (1996) VEGF-silenced breast cancer cells (Fig. 5a), or when endo- Interleukin 6 induces the expression of vascular endothelial growth – thelial cells themselves were knocked down for VEGF (data factor. J Biol Chem 271:736 741 5. Goldman S, Dirnfeld M, Koifman M, Gonen Y, Lissak A, Abramovici not shown). Interestingly, promotion of tubulogenesis by H (1993) The effect of epidermal growth factor on growth and differ- LGALS3BP was linked to galectin-3 involvement, activa- entiation of mouse preimplantation embryos in vitro. Hum Reprod tion of FAK, i.e., its phosphorylation and translocation 8:1459–1462 toward membrane sites (Fig. S3). The involvement of 6. Iacobelli S, Arno E, D'Orazio A, Coletti G (1986) Detection of antigens recognized by a novel monoclonal antibody in tissue and galectin-3 and FAK suggests that the pro-angiogenic effect serum from patients with breast cancer. Cancer Res 46:3005–3010 of LGALS3BP could be mediated by integrins, whose clus- 7. Koths K, Taylor E, Halenbeck R, Casipit C, Wang A (1993) tering is known to trigger FAK activation [29]. Galectin-3 is Cloning and characterization of a human Mac-2-binding pro- a member of the galectin family of mammalian lectins that tein, a new member of the superfamily defined by the macro- phage scavenger receptor cysteine-rich domain. J Biol Chem has affinity for B-galactoside structures [30]. Extracellular- 268:14245–14249 ly, the lectin is assumed to mediate cell–cell and cell–matrix 8. Ullrich A, Sures I, D'Egidio M, Jallal B, Powell TJ, Herbst R, interactions by binding to lactosamine-containing cell sur- Dreps A, Azam M, Rubinstein M, Natoli C et al (1994) The face glycoconjugates, including LGALS3BP. Recently, secreted tumor-associated antigen 90K is a potent immune stimu- lator. J Biol Chem 269:18401–18407 galectin-3 has been involved in VEGF/bFGF-dependent 9. Iacobelli S, Sismondi P, Giai M, D'Egidio M, Tinari N, Amatetti angiogenesis through cross-linking and clustering of integrin C, Di Stefano P, Natoli C (1994) Prognostic value of a novel αvβ3 and activation of FAK [22]. The role of galectin-3– circulating serum 90K antigen in breast cancer. Br J Cancer – LGALS3BP interaction in the angiogenic response is sub- 69:172 176 10. Tinari N, Lattanzio R, Querzoli P, Natoli C, Grassadonia A, Alberti stantiated by experiments in which this interaction is disrup- S, Hubalek M, Reimer D, Nenci I, Bruzzi P et al (2009) High ted. We found previously that synthetic lactulose amines, expression of 90K (Mac-2 BP) is associated with poor survival in which block the binding of galectins to LGALS3BP, are node-negative breast cancer patients not receiving adjuvant sys- – capable of preventing tumor cell aggregation and angiogen- temic therapies. Int J cancer J Int du cancer 124:333 338 11. Fornarini B, D'Ambrosio C, Natoli C, Tinari N, Silingardi V, esis [18]. Therefore, it could be proposed that extracellular Iacobelli S (2000) Adhesion to 90K (Mac-2 BP) as a mechanism LGALS3BP docks galectin-3 molecules, and the resulting for lymphoma drug resistance in vivo. Blood 96:3282–3285 complex cross-links and clusters the integrin on the surface 12. Strizzi L, Muraro R, Vianale G, Natoli C, Talone L, Catalano A, of the endothelial cells, thus resulting in the activation of Mutti L, Tassi G, Procopio A (2002) Expression of glycoprotein 90K in human malignant pleural mesothelioma: correlation with FAK-mediated signaling pathways that modulate the angio- patient survival. J Pathol 197:218–223 genic cascade. 13. Marchetti A, Tinari N, Buttitta F, Chella A, Angeletti CA, Sacco R, In summary, the results of this study show that Mucilli F, Ullrich A, Iacobelli S (2002) Expression of 90K (Mac-2 LGALS3BP is an important contributor to angiogenesis. BP) correlates with distant metastasis and predicts survival in stage I non-small cell lung cancer patients. Cancer Res 62:2535–2539 Anti-angiogenic intervention is receiving considerable ef- 14. Zambelli D, Zuntini M, Nardi F, Manara MC, Serra M, Landuzzi fectiveness in the suppression of cancer growth and progres- L, Lollini PL, Ferrari S, Alberghini M, Llombart-Bosch A et al sion, which includes monoclonal antibodies against (2010) Biological indicators of prognosis in Ewing's sarcoma: an angiogenic factors and synthetic small molecule inhibitors emerging role for lectin galactoside-binding soluble 3 binding protein (LGALS3BP). Int J Cancer 126:41–52 specific for membrane receptor kinases or intracellular 15. Silvestri B, Calderazzo F, Coppola V, Rosato A, Iacobelli S, Natoli mediators [23, 31]. The present findings that LGALS3BP C, Ullrich A, Sures I, Azam M, Brakebush C et al (1998) Differ- upregulates tumor VEGF and directly promotes angiogene- ential effect on TCR:CD3 stimulation of a 90-kD glycoprotein sis may hold promise for developing novel therapeutic (gp90/Mac-2BP), a member of the scavenger receptor cysteine- rich domain protein family. Clin Exp Immunol 113:394–400 agents targeting LGALS3BP that is overexpressed in a 16. Sasaki T, Brakebusch C, Engel J, Timpl R (1998) Mac-2 binding broad type of human cancers. protein is a cell-adhesive protein of the extracellular matrix which self-assembles into ring-like structures and binds beta1 integrins, collagens and fibronectin. EMBO J 17:1606–1613 17. Lee TH, Avraham HK, Jiang S, Avraham S (2003) Vascular endothelial growth factor modulates the transendothelial migration Conflict of interest None of MDA-MB-231 breast cancer cells through regulation of brain J Mol Med

microvascular endothelial cell permeability. J Biol Chem 24. Klohs WD, Hamby JM (1999) Antiangiogenic agents. Curr Opin 278:5277–5284 Biotechnol 10:544–549 18. Rabinovich GA, Cumashi A, Bianco GA, Ciavardelli D, Iurisci I, 25. Liu W, Xu J, Wang M, Wang Q, Bi Y, Han M (2011) Tumor-derived D'Egidio M, Piccolo E, Tinari N, Nifantiev N, Iacobelli S (2006) vascular endothelial growth factor (VEGF)-a facilitates tumor metas- Synthetic lactulose amines: novel class of anticancer agents that tasis through the VEGF-VEGFR1 signaling pathway. Int J Oncol induce tumor-cell apoptosis and inhibit galectin-mediated homo- 39:1213–1220 typic cell aggregation and endothelial cell morphogenesis. Glyco- 26. Jiang ZY, Chawla A, Bose A, Way M, Czech MP (2002) A biology 16:210–220 phosphatidylinositol 3-kinase-independent insulin signaling path- 19. Spinella F, Garrafa E, Di Castro V, Rosano L, Nicotra MR, Caruso way to N-WASP/Arp2/3/F-actin required for GLUT4 glucose A, Natali PG, Bagnato A (2009) Endothelin-1 stimulates lymphatic transporter recycling. J Biol Chem 277:509–515 endothelial cells and lymphatic vessels to grow and invade. Cancer 27. Zhong H, Chiles K, Feldser D, Laughner E, Hanrahan C, Georgescu Res 69:2669–2676 MM, Simons JW, Semenza GL (2000) Modulation of hypoxia- 20. Weidner N, Weinberg DS, Hardy SC, Hollister KA, Lidgard inducible factor 1alpha expression by the epidermal growth factor/ GP (1991) Localization of nuclear matrix proteins (NMPs) in phosphatidylinositol 3-kinase/PTEN/AKT/FRAP pathway in human multiple tissue types with NM-200.4 (an antibody strongly prostate cancer cells: implications for tumor angiogenesis and thera- reactive with NMPs found in breast carcinoma). Am J Pathol peutics. Cancer Res 60:1541–1545 138:1293–1298 28. Pore N, Liu S, Shu HK, Li B, Haas-Kogan D, Stokoe D, Milanini- 21. Tinari N, D'Egidio M, Iacobelli S, Bowen M, Starling G, Seachord Mongiat J, Pages G, O'Rourke DM, Bernhard E et al (2004) Sp1 is C, Darveau R, Aruffo A (1997) Identification of the tumor antigen involved in Akt-mediated induction of VEGF expression through 90K domains recognized by monoclonal antibodies SP2 and L3 an HIF-1-independent mechanism. Mol Biol Cell 15:4841–4853 and preparation and characterization of novel anti-90K monoclonal 29. Parsons JT (2003) Focal adhesion kinase: the first ten years. J Cell antibodies. Biochem Biophys Res Commun 232:367–372 Sci 116:1409–1416 22. Markowska AI, Liu FT, Panjwani N (2010) Galectin-3 is an 30. Barondes SH, Cooper DN, Gitt MA, Leffler H (1994) Galectins. important mediator of VEGF- and bFGF-mediated angiogenic Structure and function of a large family of animal lectins. J Biol response. J Exp Med 207: 1981–1993. Chem 269:20807–20810 23. Kim KJ, Li B, Winer J, Armanini M, Gillett N, Phillips HS, Ferrara 31. Fahmy RG, Dass CR, Sun LQ, Chesterman CN, Khachigian LM N (1993) Inhibition of vascular endothelial growth factor-induced (2003) Transcription factor Egr-1 supports FGF-dependent angio- angiogenesis suppresses tumour growth in vivo. Nature 362:841– genesis during neovascularization and tumor growth. Nat Med 844 9:1026–1032